In many respects, then, it casts the theory of public policy making as a theory of
learning.
A belief system is organized in three tiers: what Sabatier terms a ‘‘deep core’’ of
normative belief or ideology which can be expected to hold across domains; a
‘‘policy core’’ of more speciWc commitments within a domain; and then non-
essential or secondary matters of detail. What holds a coalition together is agree-
ment over a policy core, and the only way this core can change is as an eVect of
some external and fundamental shock. Within a domain, however, learning takes
place between coalitions as a result of diVerences in their belief systems. The
likelihood of learning is inversely related to the level of commitment to a belief,
such that secondary aspects of a policy or program are more likely to be revised or
amended in the light of new evidence than elements of the policy core. The process
of learning is facilitated by the existence of a professional forum in which members
of diVerent coalitions may exchange views and interpretations of both problems
and solutions.
Frank Fischer ( 2003 ) presents a social constructionist critique of the advocacy
coalition framework, drawing on Maarten Hajer’s work on discourse coalitions
(Hajer 1995 ). His argument is that belief systems are not pre-existing and empirically
veriWable in the way Sabatier and colleagues might claim, but are instead better
understood as narratives or storylines. A common interpretation of a problem and
appropriate solutions to it is not the basis for membership of a coalition, but
something which its various members produce together, through their communica-
tions and interactions. Indeed, a common storyline is likely to be more powerful and
eVective the more it is susceptible to a variety of interpretations.
3.2 Social Learning
Peter Hall’s inXuential treatment of what he calls ‘‘social learning’’ is based on a
study of economic policy making in Britain in the 1970 s and 1980 s (Hall 1993 ). 9 He
is interested in the ‘‘interpretative framework’’ of policy, meaning the common
understanding of its goals and instruments as well as the nature of the problems to
which policy is addressed. Drawing on Kuhn ( 1962 ), he refers to this as a ‘‘paradigm,’’
and the question he asks is why it changes or shifts, that is, how and why a
policy community learns to think diVerently. For what is at issue in Heclo’s largely
technocratic model of policy learning is the idea of the relative autonomy of the
state from societal pressure. Is ‘‘learning’’ really conWned to a ministerial and
administrative elite?
9 Hall’s work has inspired and inXuenced a small literature on macroeconomic policy learning in the
UK, as James and Lodge ( 2003 ) point out.
374 richard freeman